On Tue, Dec 20, 2016 at 06:15:57PM +0200, Eli Zaretskii wrote:
> > Date: Tue, 20 Dec 2016 12:18:58 +0100
> > From: "Dr. Werner Fink" <werner@suse.de>
> > Cc: martin rudalics <rudalics@gmx.at>, 25228@debbugs.gnu.org
> > 
> > > Can you describe how that mechanism is triggered by the offending
> > > change?
> > 
> > Just play around with eval-region/expression ... that is that I can switch with
> > 
> >   (font-setting-change-default-font ":0" t)
> > 
> > or
> > 
> >   (font-setting-change-default-font ":0" nil)
> > 
> > to the system default font settings and then with
> > 
> >   (custom-set-faces '(default ((t (:family "B&H LucidaTypewriter"
> >     :foundry "B&H" :slant normal :weight normal :height 110 :width normal)))))
> 
> Thanks.  However, I don't have easy access to systems where this
> problem happens, so I must examine the source code instead.  That is
> why I asked to describe which code triggers this and how it started
> affecting us after commit e65c307.

The only guess I have is that the correction had now enabled the :user-spec
key/attribute.  For a workaround I've change some lines of code in
lisp/dynamic-setting.el ... compare with the attached patch.

With this patch both the custom font if enabled in ~/.emacs as well as
the system font works here for GNU Emacs with GTK as well as with X11.


Werner

-- 
  "Having a smoking section in a restaurant is like having
          a peeing section in a swimming pool." -- Edward Burr